Nobody has to take notes again.
It joins the meeting, writes down who said what, and then reads the transcript back to you: who did the talking, who asked the questions, where it got heated, and what people promised out loud. Then it turns the agreements into tasks with owners.
Fifty hours of meetings included, and they cost no AI credits.
It writes down every turn, then reads the transcript back: talk share per person, questions asked against questions answered, and the commitments somebody made out loud.
What it solves
Four problems this solves.
The first three are about the meeting you are in. The fourth is about all the meetings you have already recorded.
Whoever writes gets a worse meeting, because they are typing instead of listening. So either the notes are thin or the best person in the room is half absent.
A week later two people are certain of two different things, and there is no record to settle it. Usually neither of them is lying.
You know some meetings run badly. You cannot say whether one person dominates every call, or whether the habit you tried to fix in March actually changed.
Anything that needs AI has to run on each meeting separately, so your older recordings stay a pile nobody opens again.
The read
See who talked, who asked and what was agreed.
Talk share per person, questions asked against questions answered, how the room moved from opening to close, who interrupted whom, and the words that kept coming back.
| Person | Turns | Questions | Longest run | Share of the talking |
|---|---|---|---|---|
| Dana | 41 | 14 | 190 words | 44% |
| Rami | 33 | 6 | 240 words | 33% |
| Priya | 22 | 9 | 110 words | 18% |
| Omar | 7 | 1 | 80 words | 5% |
Why it is not AI
The analysis does not use AI, so it works on every meeting you already have.
Everything on that screen is counted from the words, not inferred by a model. That gives you three things a model-based tool cannot.
What that buys you
- It works on your back catalogue. Every meeting you have ever recorded can be read tonight, not just the ones from tomorrow.
- It costs nothing. No credits, no per-meeting charge, no cap to watch.
- It says the same thing twice. Open it next quarter and the numbers have not moved, so you can argue from them.
- It cannot make anything up. There is no summary to be subtly wrong about.
How each number is worked out
- Talk share and turns. It counts the words each person said and divides. There is nothing to get wrong.
- Questions asked and answered. A question mark makes a question, and the next turn is whoever answered it.
- Interruptions. A turn that ends mid-sentence with somebody else starting. It counts them and leaves the judgement to you.
- Commitments and figures. Phrases like "I will" and "by Friday", plus any amounts anybody quoted.
The half that genuinely does use a model isCoach, which is about how you ran the meeting rather than what happened in it. It is labelled as AI everywhere it appears.
What it connects to
What was agreed becomes tasks on the right records.
The recording is the start of a chain, not the end of one, which is only possible because the calendar, the deal and your task list are already here.
Where it reaches
- Calendar. It joins from the meeting on your calendar, so nobody has to remember to start a recorder.
- Sales. A discovery call is measured against what a discovery call is supposed to establish, and the answers land on the deal.
- Workspace. What the meeting agreed becomes tasks with owners, in the same list as everything else you owe.
- NearSync AI. Coaching is the one half that runs on a model, on whichever one you have configured.
What you get afterwards
- A written summary of what was said and settled.
- Tasks with owners, from the things people committed to out loud.
- Answers on the deal, where the meeting type asked for them.
- A coaching note, if you want one.
Meeting types
A discovery call and a retro are not the same meeting.
Each type carries its own process, so what gets measured afterwards is what that kind of meeting was supposed to achieve rather than one generic score.
Name it, file it under a department, and describe it in a sentence. The description is not decoration: it goes into the analysis, so "a first call to establish whether there is a real opportunity" reads differently from "sales call".
What you write here changes what the analysis looks for.Saved, the analysis instruction is composed behind the scenes from the framework, your list and the sentence you wrote. Every meeting of this kind is then read the same way, which is what makes one quarter comparable with the next.
The frameworks
Twenty-five of them, and you can see what each one looks for.
Nothing is scored against a rubric you are not allowed to read. Open a framework and it tells you what it measures and why that matters for that meeting.
Budget, Authority, Need, Timeline
A deeper checklist for complex deals
Situation, Problem, Implication, Need-Payoff
Pain, Budget, Decision
Start from a blank list and write your own.
You see your own department's frameworks plus the general ones, so nobody scrolls past twenty-five to find the four that apply to them.
Every capability
Four things it does.
Two for getting the meeting in, two for reading it back. Only one of the four runs on a model, and it is the one marked.
Capturing it
The questions we get asked.
Anything longer lives in the help centre.
How many hours of meetings can we record?
50 hours a month on Standard and 100 on Pro, across the whole company. Recording, transcript, summary and coaching are all included in that.
Which meeting platforms does it join?
Google Meet, Microsoft Teams and NearSync Huddle. It joins on its own, so nobody has to remember to press record.
What does it give us afterwards?
Who said what, who committed to what, and the tasks that came out of it, filed against the deal or the project they belong to.
Does it tell us how the meeting went?
Yes. Twenty-five frameworks measure how a meeting was run, and the coaching is honest about it rather than flattering.
Bring a recording of a real meeting.
Half an hour. The analysis runs on it there and then, because it needs nothing set up first.